1: %\documentclass[aps,preprint,showpacs]{revtex4}
2: %\documentclass[floats,aps,showpacs]{revtex4}
3: \documentclass[floats,aps,showpacs,twocolumn]{revtex4}
4:
5: \usepackage[dvips]{graphicx}
6: \usepackage{amsfonts}
7:
8:
9:
10:
11:
12: \newcommand{\heff}{h_{\text{eff}}}
13: \newcommand{\Areg}{A_{\text{reg}}}
14: \newcommand{\Wch}{W_{\text{ch}}}
15: \newcommand{\freg}{f_{\text{reg}}}
16: \newcommand{\tH}{\tau_{\text{\tiny{H}}}}
17: \newcommand{\Dch}{\Delta_{\text{ch}}}
18: \newcommand{\mmax}{m_{\text{max}}}
19: \newcommand{\mstar}{m^{*}}
20:
21: \newcommand{\PSImagx}[2]{\includegraphics[width=#2]{#1}}
22:
23: \newcommand{\Z}{\mathbb{Z}}
24:
25:
26: \begin{document}
27:
28: \title{Flooding of Chaotic Eigenstates into Regular Phase Space Islands}
29:
30: \author{Arnd B\"acker, Roland Ketzmerick, and Alejandro G. Monastra}
31:
32: \affiliation{Institut f\"ur Theoretische Physik, Technische
33: Universit\"at Dresden, 01062 Dresden, Germany}
34:
35: \date{\today}
36:
37: \begin{abstract}
38: We introduce a criterion for the existence of regular states in
39: systems with a mixed phase space. If this condition is not fulfilled
40: chaotic eigenstates substantially extend into a regular island. Wave
41: packets started in the chaotic sea progressively flood the island. The
42: extent of flooding by eigenstates and wave packets increases
43: logarithmically with the size of the chaotic sea and the time,
44: respectively. This new effect is observed for the example of
45: island chains with just 10 islands.
46: \end{abstract}
47: \pacs{05.45.Mt, 03.65.Sq}
48:
49: \maketitle
50:
51:
52:
53: One of the cornerstones in the understanding of the structure of
54: eigenstates in quantum systems is the semiclassical eigenfunction
55: hypothesis\ \cite{Scefct}: in the semiclassical limit the eigenstates
56: concentrate on those regions in phase space which a typical orbit
57: explores in the long-time limit. For integrable systems these are the
58: invariant tori. For ergodic dynamics the eigenstates become
59: equidistributed on the energy shell \cite{Qerg}. Typical systems have
60: a mixed phase space, where regular islands and chaotic regions
61: coexist. In this case the semiclassical eigenfunction hypothesis
62: implies that the eigenstates can be classified as being either regular
63: or chaotic according to the phase-space region on which they
64: concentrate. Note, that this may fail for an infinite phase space
65: \cite{HufKetOttSch2002}.
66:
67:
68: In this paper we study mixed systems with a compact phase space, but
69: away from the semiclassical limit. Here the properties of eigenstates
70: depend on the size of phase-space structures compared to Planck's
71: constant $h$. In the case of 2D maps this can be very simply stated
72: \cite{BerBalTabVor79}: a regular state with quantum number $m=0,1,...$
73: will concentrate on a torus enclosing an area $(m+1/2) h$, as can be
74: seen in Fig.~\ref{fig:Husimis}(c).
75:
76: We will show that this WKB-type quantization rule is not a sufficient
77: condition. We find a second criterion for the existence of a regular
78: state on the $m$-th quantized torus,
79: \begin{equation} \label{newcondition}
80: \gamma_m < \frac{1}{\tH} .
81: \end{equation}
82: Here $\tH = h / \Dch$ is the Heisenberg time of the chaotic sea with
83: mean level spacing $\Dch$ and $\gamma_m$ is the decay rate of the
84: regular state $m$ if the chaotic sea were infinite. Quantized tori
85: violating this condition will not support regular states. Instead,
86: chaotic states will {\it flood} these regions, see
87: Fig.~\ref{fig:Husimis}(a). In terms of dynamics we find that wave
88: packets started in the chaotic sea progressively flood the island as
89: time evolves. Partial and even complete flooding is possible,
90: depending on system properties. These findings are relevant for
91: islands surrounded by a large chaotic sea.
92:
93: We numerically demonstrate the flooding and the disappearance of
94: regular states for the important case of island chains. In typical
95: Hamiltonian systems they appear around any regular island. On larger
96: scales they are relevant for Hamiltonian ratchets
97: \cite{SchOttKetDit2001}, the kicked rotor with accelerator modes
98: \cite{Izr90}, and the experimentally
99: \cite{SteOskRai2001,Hen2001,expAccel} and theoretically
100: \cite{theoryAccel} studied kicked atom systems. The flooding of
101: regular islands by chaotic states is a new quantum signature of a
102: classically mixed phase space. This phenomenon shows that not only
103: local phase-space structures, but also global properties of the phase
104: space determine the characteristics of quantum states.
105:
106: \begin{figure}[b]
107: \begin{center}
108: \PSImagx{fig1.eps}{8.6cm}
109: \caption{(a) Husimi representation of a chaotic state flooding the
110: regular islands. Shown are three phase-space cells out of
111: $M=1597$. Full curves are classical tori close to the border of
112: the regular islands and black dots are iterates of a chaotic
113: orbit. This eigenstate extends well into the islands, while having
114: no weight in their central region. (b) Average of the eigenstate
115: over all $M$ cells. (c) For $M=1$ eigenstates concentrate
116: either on the chaotic component (left) or over the $m$-th
117: quantized regular torus. For all plots $\heff \approx 1/30$ is
118: used.
119: \label{fig:Husimis} } \end{center}
120: \end{figure}
121:
122: Before we explain the origin of Eq.~(\ref{newcondition}), we
123: numerically study its consequences. We choose a system, where we can
124: change $\tH$ by increasing the system size without affecting the rates
125: $\gamma_m$. A one dimensional kicked system
126: \begin{equation}
127: H(p,x,t) = T(p) + V(x) \sum_n \delta(t-n) , \label{hamiltonian}
128: \end{equation}
129: has a stroboscopic time-evolution given by the mapping, $x_{t+1}= x_t
130: + T'(p_t), \ p_{t+1} = p_t-V'(x_{t+1})$. The phase space is compact with
131: periodic boundary conditions for $x_t\in [0,M]$ and
132: $p_t\in[-1/2,1/2]$. Choosing the functions $V'(x)$ and $T'(p)$
133: appropriately \cite{fctchoice} we get a chain of $M$ islands, one per
134: unit cell (see Fig.~\ref{fig:Husimis}). The islands cover a relative area
135: $\Areg=0.215$ and have fine structure close to their boundary that is
136: negligible for the quantum properties studied here. Points inside an
137: island are mapped one unit cell to the right, i.e.\ the island chain
138: is transporting.
139:
140: The eigenstates $|\psi\rangle$ of the quantum system are determined by
141: the eigenvalue equation, $\hat{U}|\psi\rangle = e^{i \varphi}
142: |\psi\rangle$, where $\hat U$ is the unitary time-evolution operator
143: over one time period,
144: $\hat{U}= \exp[-2\pi i T(\hat{p})/\heff]\,\exp[-2\pi i V(\hat{x})/\heff]$.
145: The spatial periodicity after $M$
146: cells requires an effective Planck's constant $\heff=M/N$, with
147: incommensurate integers $M$ and $N$. We choose for $M/N$ the rational
148: approximants of $\heff=1/(d+\sigma)$ with $\sigma=(\sqrt{5}-1)/2$ the
149: golden mean and, e.g., $d=29$ in Fig.~\ref{fig:Husimis}. This ensures
150: that there are no undesired periodicities and that $\heff$ is
151: approximately constant when varying $M$. Moreover, the
152: operator $\hat{U}$ reduces to an $N\times N$ periodic band
153: matrix. Using the symmetrized version of the map and making a unitary
154: transformation to a band matrix we are able to calculate the
155: eigenstates $|\psi\rangle$ of $\hat{U}$ up to $N \approx
156: 10^5$.
157:
158: For $M=1$, Fig.~\ref{fig:Husimis}(c) shows a typical chaotic
159: eigenstate and five regular states. The chaotic state extends over
160: the chaotic phase-space component and the regular states concentrate
161: on quantized tori. The eigenstates are represented on the classical
162: phase space by the Husimi distribution, where for visualization we use
163: tilted coherent states adapted to the shape of the island. For larger
164: system sizes we find that chaotic states flood the islands of
165: classically regular motion. Fig.~\ref{fig:Husimis}(a) shows such a
166: state for $M=1597$ that clearly ignores the outer tori of the island,
167: which for $M=1$ act as barriers for chaotic states
168: (Fig.~\ref{fig:Husimis}(c), left). In the central part of the island,
169: however, this state has essentially no weight. This partial flooding
170: of the island is observed even better in Fig.~\ref{fig:Husimis}(b),
171: where an average of the Husimi function of this state is taken over
172: all $M$ unit cells. The almost constant value in the chaotic component
173: extends well into the island. Inside the island, clearly away from
174: its outer boundary, the Husimi function sharply drops to zero.
175:
176: \begin{figure}[bth]
177: \begin{center}
178: \PSImagx{fig2.eps}{8.2cm}
179: \caption{Distribution $P(W)$ of the weights $W$ of the
180: eigenstates in the regular islands for $\heff \approx 1/10$. (a)
181: $M=1$: The main peak near $W=0$ is due to chaotic states and two
182: further peaks are due to regular states $m=0, 1$ (see
183: insets). (b) For increasing $M$ the main peak shifts to larger
184: values of $W$ (white line) and the two peaks from the regular
185: states disappear sequentially. (c) $M=10946$: A narrow peak
186: remains around $W = \Areg$. Three phase space cells of an
187: eigenstate show a complete flooding of the islands.
188: \label{fig:weight-distribution}}
189: \end{center}
190: \end{figure}
191:
192: For a quantitative description of this flooding we now analyze the
193: weight $W$ of each eigenstate inside the islands. We determine this
194: weight by integrating the normalized Husimi function (calculated on a
195: $30 \times 30$ grid per unit cell) over the islands. In the
196: semiclassical limit, $\heff\to 0$, regular states have $W=1$, while
197: chaotic states have $W=0$. The distribution $P(W)$ of these weights
198: for all eigenstates is shown in Fig.~\ref{fig:weight-distribution} for
199: various system sizes and $\heff \approx 1/10$ \cite{footnote1}. For
200: $M=1$ we observe, as expected, a main peak near $W=0$ coming from the
201: chaotic eigenstates and two distinct peaks at larger $W$ from the two
202: regular states, see Fig.~\ref{fig:weight-distribution}(a). A
203: remarkable shift of the main peak of $P(W)$ to larger values of $W$
204: can be observed in Fig.~\ref{fig:weight-distribution}(b). This shows
205: that by enlarging the system size $M$ all chaotic states continuously
206: increase their weight inside the regular islands. This increase stops
207: when the center of the main peak reaches $W = 0.22$, which corresponds
208: to the area $\Areg$ of the island. For these system sizes all states
209: completely flood the island (Fig.~\ref{fig:weight-distribution}(c),
210: inset), as observed in Ref.~\cite{HufKetOttSch2002}.
211:
212: What happens to the regular states as $M$ is increased?
213: Figure~\ref{fig:weight-distribution}(b) shows that the corresponding
214: peaks in the distribution $P(W)$ disappear. Notably, the peak for
215: $m=0$ is much longer visible than the peak for $m=1$. The nemesis of
216: the regular states can be quantified by determining their fraction
217: $\freg$ as a function of $M$. To this end we define a state to be
218: regular when its weight $W$ inside the islands exceeds 50\%, where the
219: exact criterion does not affect our analysis.
220: Figure~\ref{fig:fractions} shows that the fraction $\freg$ decreases
221: from approximately $\Areg$ all the way to zero. The decay is slower
222: for smaller $\heff$.
223:
224: \begin{figure}[t]
225: \begin{center}
226: \PSImagx{fig3.eps}{8.6cm}
227: \caption{Fraction $\freg$ of regular states vs.\ system size $M$
228: for $\heff\approx 1/5$ (triangles), $1/10$ (circles), and $1/20$
229: (squares). An approximately linear decrease with $\ln M$ to
230: $\freg=0$ can be seen (lines are a guide to the eye). Already for
231: small system sizes $M \approx 10$ and small $\heff$ a significant
232: reduction of regular states is observed.
233: \label{fig:fractions} } \end{center}
234: \end{figure}
235:
236: Remarkably, in Fig.~\ref{fig:fractions} we see
237: strong signatures of the decrease of $\freg$ already for small system
238: sizes $M \approx 10$. This holds even for small values of $\heff$,
239: where the complete flooding of the island is numerically not
240: accessible. Similarly, a shift of the main peak in
241: Fig.~\ref{fig:weight-distribution}(b) is clearly detectable for such
242: small systems. We thus find that partial flooding of regular islands
243: is easily observable.
244:
245: Why do chaotic states flood the islands of regular motion and why do
246: regular states disappear as the system size is increased? Let us
247: consider a single regular island coupled by tunneling to a chaotic
248: sea. If the chaotic sea is infinite, its states form a continuum. A
249: regular state on the $m$-th quantized torus has a decay rate
250: $\gamma_m$ to the continuum \cite{footnotedecay}. Thus, it is not an
251: eigenstate, but it is dissolved into chaotic states. As a consequence,
252: the chaotic states occupy the $m$-th quantized torus of the island. If
253: the chaotic sea is finite, but large enough, this decay of the $m$-th
254: regular state may still take place. The condition for the decay is
255: that during the time $1/\gamma_m$ the discrete chaotic spectrum is not
256: resolved, leading to $1/\gamma_m \ll h/\Dch = \tH$ \cite{FGR}. On the
257: other hand, if the chaotic sea is so small that during the time
258: $1/\gamma_m$ the chaotic spectrum is well resolved, then the regular
259: state $m$ does not decay, yielding Eq.~(\ref{newcondition}). Note,
260: that $\gamma_m$ increases monotonically with $m$, as for larger $m$
261: the $m$-th torus is closer to the boundary of the island.
262:
263: The quantized tori of an island can thus be grouped into two classes:
264: (i) The innner tori, $m=0,...,\mstar-1$, where condition
265: (\ref{newcondition}) is fulfilled and regular states exist. (ii) The
266: outer tori, $m=\mstar, ...,\mmax-1$, where Eq.~(\ref{newcondition}) is
267: violated and which is flooded by chaotic states. Here $\mmax$ is the
268: number of quantized tori at a given $\heff$. We find for the fraction
269: $\freg$ of regular states and the weight $\Wch$ of chaotic states
270: inside the island:
271: \begin{equation} \label{eq:freg-Wch}
272: \freg=\Areg \frac{\mstar}{\mmax} , \quad
273: \Wch= \Areg \left( 1- \frac{\mstar}{\mmax} \right) .
274: \end{equation}
275:
276: Variation of the system size $M$ in our example allows to change
277: the Heisenberg time $\tH \sim M$,
278: while keeping the rates $\gamma_m$ fixed.
279: Enlarging $M$ leads via Eq.~(\ref{newcondition})
280: to a decrease of $\mstar$,
281: starting from $\mstar=\mmax-1$ all the way to $\mstar=0$.
282: Together with Eq.~(\ref{eq:freg-Wch})
283: this explains Fig.~\ref{fig:weight-distribution}(b),
284: where the regular state with $m=1$ disappears
285: before the $m=0$ state
286: and the weight $\Wch$ grows until $\Wch=\Areg$,
287: where the island is completely flooded,
288: see Fig.~\ref{fig:weight-distribution}(c).
289: This also explains the decrease of $\freg$ from $\Areg$ to $0$,
290: as observed in Fig.~\ref{fig:fractions}.
291: This decrease occurs over an exponentially large range
292: in $M$, due to the roughly
293: exponential dependence of $\gamma_m$ on $m$.
294: A quantitative understanding requires a theory
295: for the decay rates $\gamma_m$,
296: which is a subject of current research on dynamical tunneling
297: \cite{DynTun,PodNar2003}.
298: Note, that in the case of chaos-assisted tunneling
299: the splitting of symmetry related regular
300: states fluctuates strongly, depending on individual chaotic states.
301: In contrast, the decay rate $\gamma_m$ describes an average
302: tunneling to a continuum of chaotic states.
303:
304: Variation of $\heff$ affects both $\gamma_m$ and $\tH$ in
305: Eq.~(\ref{newcondition}). While $\tH\sim M/\heff$,
306: one expects in analogy to WKB theory that
307: $\gamma_m \sim \exp[ - g(m/\mmax)/\heff]$,
308: where the system specific function $g$ decreases monotonically to $g(1)=0$.
309: From the definition of $\mstar$ follows
310: $\mstar/\mmax = g^{-1}[\heff \ln(M/\heff)]$,
311: where $g^{-1}$ decreases monotonically.
312: Decreasing $\heff$ reduces the argument of $g^{-1}$
313: such that $\mstar/\mmax$ increases.
314: Eq.~(\ref{eq:freg-Wch}) implies that
315: $\freg$ grows and $\Wch$ decreases.
316: Note that in the semiclassical limit, $\heff\to 0$,
317: we obtain $\mstar/\mmax\to 1$.
318: This is in agreement with the semiclassical eigenfunction
319: hypothesis, namely $\freg=\Areg$
320: and there is no flooding.
321: In contrast, if the system size is infinite, we have an infinite
322: $\tH$ and our argument leads to $\mstar=0$, i.e.\ complete flooding,
323: for any $\heff$. This coincides with the considerations of
324: \cite{HufKetOttSch2002}
325: implying a failure of the semiclassical eigenfunction hypothesis.
326:
327:
328: Our explanation is complete for systems without localization. For
329: example, this is the case if the average classical drift of a unit
330: cell is non-zero, like in atom optic experiments in the presence of
331: gravity ~\cite{expAccel,theoryAccel}. Localization, however, sets a
332: lower bound to the effective mean level spacing, $\Dch \sim
333: 1/\lambda$, where $\lambda$ is the localization length. For $M >
334: \lambda$, this leads to $\tH \sim \lambda$ and $\mstar$ stays at its
335: value for $M=\lambda$. According to Eq.~(\ref{eq:freg-Wch}) the same
336: holds for $\freg$ and $\Wch$. This applies, e.g., to dynamical
337: localization in the kicked rotor. For transporting islands, like in
338: the model studied here, $\lambda \sim 1/\gamma_0$ is unusually large
339: \cite{HanOttAnt1984,IomFisZas2002andrefs,HufKetOttSch2002},
340: such that already for $M = \lambda$ one has $\mstar=0$, $\freg=0$, and
341: $\Wch=\Areg$. In this case, localization has no consequences
342: \cite{footnote2}.
343:
344:
345:
346:
347:
348:
349:
350: \begin{figure}[tbp]
351: \begin{center}
352: \PSImagx{fig4.eps}{8.0cm}
353: \caption{Weight $W$ in the island vs.\ time~$t$ for a wave packet
354: initially started as a coherent state in the chaotic sea at
355: $(x,p)=(0.4,-0.2)$ for system sizes $M=1,13,144,1597$ and
356: $\heff\approx 1/30$. The thick (green) line guides the eye to
357: the linear increase with $\ln t$ until $W=\Areg$. Insets show the
358: time-evolved wave packet averaged over all cells for the case
359: $M=1597$, demonstrating the progressive flooding of the island.
360: \label{fig:temp-flooding} }
361: \end{center}
362: \end{figure}
363:
364:
365:
366:
367:
368:
369: We generalize our analysis to the dynamics of wave packets, which is
370: experimentally of great relevance~\cite{expAccel}. A wave packet
371: started on the $m$-th torus will be restricted to that region if
372: condition (\ref{newcondition}) is fulfilled, i.e.\ $m<\mstar$. If
373: $m>\mstar$, however, the wave packet will decay into the chaotic sea.
374: Particularly interesting is the case of a wave packet started inside
375: the chaotic sea. The island is progressively flooded, i.e.\ the $m$-th
376: torus at time $t_m \sim 1/\gamma_m$ for $m>\mstar$. For $t>\tH$ the
377: weight $W(t)$ will saturate at $\Wch$, Eq.~(\ref{eq:freg-Wch}). This
378: is confirmed in Fig.~\ref{fig:temp-flooding}, for increasing values of
379: $M$.
380:
381: Our results have consequences for spectral statistics in mixed systems
382: which go well beyond the previously studied effects of dynamical
383: tunneling (see e.g.\ \cite{ProRob94b,PodNar2003}). The effective size
384: of the regular region, $\freg$ in Eq.~(\ref{eq:freg-Wch}), entering
385: the Berry-Robnik formula \cite{BerRob84} is drastically reduced.
386:
387:
388: Our analysis applies as well to hierarchical
389: states~\cite{KetHufSteWei2000}, which are confined by partial
390: transport barriers with turnstile areas smaller than $h$. We predict
391: the additional condition $\gamma < 1/\tH$ for their existence, where
392: $\gamma$ describes the decay through these partial barriers. For
393: regular states on island chains within that hierarchical region,
394: condition (\ref{newcondition}) applies, with $\tH$ given by the mean
395: level spacing of the surrounding hierarchical states.
396:
397: Finally, we emphasize that the time periodicity of the system
398: (\ref{hamiltonian}) and the restriction of our discussion to maps is
399: not crucial and that we expect flooding of islands for any Hamiltonian
400: with a mixed phase space. We stress that this new quantum signature of
401: chaos for eigenstates and wave packet dynamics already appears for
402: small system sizes, e.g.\ island chains of length 10. This makes
403: numerical explorations very feasible and should lead to experimental
404: observations, for example using optical
405: lattices~\cite{SteOskRai2001,Hen2001,expAccel}.
406:
407: We thank Holger Schanz for discussions and the DFG for support under
408: contract KE 537/3-2.
409:
410:
411: \begin{thebibliography}{10}
412:
413: \bibitem{Scefct}
414: I.~C. Percival, J. Phys. B {\bf 6}, L229 (1973); M.~V. Berry,
415: J. Phys. A {\bf 10}, 2083 (1977); A.~Voros, in {\it Stochastic
416: Behavior in Classical and Quantum Hamiltonian Systems} (Springer,
417: Berlin, 1979), pp. 326.
418:
419: \bibitem{Qerg}
420: See e.g.\ A.~B\"acker, R.~Schubert and P.~Stifter, Phys. Rev. E {\bf
421: 57}, 5425 (1998), and references therein.
422:
423: \bibitem{HufKetOttSch2002}
424: L. Hufnagel, R. Ketzmerick, M.-F. Otto, and H. Schanz,
425: Phys. Rev. Lett. {\bf 89}, 154101 (2002).
426:
427: \bibitem{BerBalTabVor79}
428: M.~V. Berry, N.~L. Balazs, M. Tabor, and A. Voros, Ann. Phys. {\bf
429: 122}, 26 (1979).
430:
431: \bibitem{SchOttKetDit2001}
432: H. Schanz, M.-F. Otto, R. Ketzmerick, and T. Dittrich,
433: Phys. Rev. Lett. {\bf 87}, 070601 (2001).
434:
435: \bibitem{Izr90}
436: F.~M. Izrailev, Phys. Rep. {\bf 196}, 299 (1990).
437:
438: \bibitem{SteOskRai2001}
439: D.~A. Steck, W.~H. Oskay, and M.~G. Raizen, Science {\bf 293}, 274
440: (2001).
441:
442: \bibitem{Hen2001}
443: W.~K. Hensinger {\it et~al.}, Nature {\bf 412}, 52 (2001).
444:
445: \bibitem{expAccel}
446: M.~K. Oberthaler {\it et~al.}, Phys. Rev. Lett. {\bf 83}, 4447 (1999);
447: S. Schlunk, M.~B. d'Arcy, S.~A. Gardiner, and G.~S. Summy,
448: Phys. Rev. Lett. {\bf 90}, 124102 (2003).
449:
450: \bibitem{theoryAccel}
451: S. Fishman, I. Guarneri, and L. Rebuzzini, Phys. Rev. Lett. {\bf 89},
452: 084101 (2002), J. Stat. Phys. {\bf 110}, 911 (2003); M.~B. d'Arcy {\it
453: et~al.}, Phys. Rev. E {\bf 69}, 027201 (2004).
454:
455: \bibitem{fctchoice}
456: We start with linear functions $t'(p)=1/2\pm(1-2p)$ for $0<\pm p<1/2$
457: (periodically extended) and $v'(x) = -k-r(x-k)$ for $k-1/2<x<k+1/2$,
458: where $k\in\Z$. Smoothing with a Gaussian,
459: $G(z)=\exp(-z^2/2\varepsilon^2)/ \sqrt{2\pi\varepsilon^2}$, gives
460: analytic functions $T'(p) = \int dz \; t'(p+z) \; G(z)$ and $V'(x) =
461: \int dz \; v'(x+z) \; G(z)$. We use $r=0.65$, and $\varepsilon=0.015$,
462: as for Fig.~1a in \cite{HufKetOttSch2002}.
463:
464: \bibitem{footnote1}
465: We improve the statistics for small $M$ by varying the phase of the
466: boundary condition in the $x$-direction, which does not affect the
467: underlying classical dynamics. For large $M$ we use 2000 eigenstates.
468:
469: \bibitem{footnotedecay}
470: This is in analogy to a discrete state coupled to a continuum
471: discussed in [U. Fano, Nuovo Cimento, N. s. {\bf 12}, 154 (1935)]. In
472: practice one can determine the rates $\gamma_m$ by starting suitable
473: wave packets as in \cite{HanOttAnt1984}.
474:
475: \bibitem{HanOttAnt1984}
476: J.~D. Hanson, E. Ott, and T.~M. Antonsen, Phys. Rev. A {\bf 29}, 819 (1984).
477:
478: \bibitem{FGR}
479: A barrier tunneling analogy is given in E. J. Heller and M. J. Davies,
480: J. Phys. Chem {\bf 85}, 307 (1981). For a discussion of time scales in
481: quantum chaos see e.g.\ [F.~Borgonovi and G.~Casati, in {\it Frontiers
482: in Quantum Physics} (Springer, Singapore, 1998), pp. 127,
483: cond-mat/9711281]. By Fermi's golden rule one can show that the
484: reasoning using time scales is equivalent to the comparison of the
485: coupling matrix element $v_m$ to the mean spacing $\Dch$.
486:
487: \bibitem{PodNar2003}
488: V.~A. Podolskiy and E.~E. Narimanov, Phys. Rev. Lett. {\bf 91}, 263601
489: (2003).
490:
491: \bibitem{DynTun}
492: M.~J. Davis and E.~J. Heller, J. Chem. Phys. {\bf 75}, 246 (1981);
493: O. Bohigas, S. Tomsovic, and D. Ullmo, Phys. Rep. {\bf 223}, 43
494: (1993); O. Brodier, P. Schlagheck, and D. Ullmo, Ann. Phys. {\bf
495: 300}, 88 (2002); C. Eltschka and P. Schlagheck, Phys. Rev. Lett. {\bf
496: 94}, 014101 (2005).
497:
498: \bibitem{IomFisZas2002andrefs}
499: A. Iomin, S. Fishman, and G.~M. Zaslavsky, Phys. Rev. E {\bf 65},
500: 036215 (2002) and references therein.
501:
502: \bibitem{footnote2}
503: Depending on the factor between $\lambda$ and $1/\gamma_0$, the $m=0$
504: state might couple to very few chaotic states only. The emerging
505: eigenstates may look more like hybrid states that are regular in some
506: unit cells and chaotic in others, explaining such an observation in
507: Fig.~4 of Ref.~\cite{HufKetOttSch2002}.
508:
509: \bibitem{ProRob94b}
510: T. Prosen and M. Robnik, J. Phys. A {\bf 27}, 8059 (1994).
511:
512: \bibitem{BerRob84}
513: M.~V. Berry and M. Robnik, J. Phys. A {\bf 17}, 2413 (1984).
514:
515: \bibitem{KetHufSteWei2000}
516: R. Ketzmerick, L. Hufnagel, F. Steinbach, and M. Weiss,
517: Phys. Rev. Lett. {\bf 85}, 1214 (2000).
518:
519: \end{thebibliography}
520:
521:
522: \end{document}
523: